AI-driven Surrogate Modeling for Iced Tailplane Aerodynamic Prediction: Matthew J. Roberts, Philip Cornford, Derya Tilki. Oncological Benefits of Extended Pelvic Lymph Node Dissection: More Fog or Clarity to the Debate? Eur Urol. In press. https://doi.org/10.1016/j.eururo.2024.12.001

Aircraft tailplane icing poses a significant challenge to flight safety and aerodynamic efficiency, particularly under holding or approach conditions. While high-fidelity numerical tools exist for simulating ice accretion and its aerodynamic effects, their computational cost is prohibitive for integration into design workflows such as multidisciplinary design optimization (MDO). This work introduces a hybrid surrogate modeling framework that combines physics-based modeling with machine learning techniques to enable rapid aerodynamic assessment of iced configurations. An in-house dataset of 2D iced airfoils is generated using a numerical icing code and RANS-based CFD simulations. Neural networks, namely Multi-Layer Perceptrons (MLPs) for iced airfoil geometry prediction and Convolutional Neural Networks (CNNs) with Signed Distance Fields (SDFs) for lift prediction, are trained to replace expensive CFD computations. These models are then integrated into a quasi-3D nonlinear Vortex Lattice Method (VLM), allowing spanwise interpolation and global aerodynamic prediction for clean and iced tailplane geometries. The proposed framework achieves a three to four order-of-magnitude reduction in computational cost while preserving key aerodynamic trends with acceptable accuracy. A comparative study against full RANS simulations demonstrates good agreement, indicating the method’s suitability for use in early phase MDO of ice tolerant tailplanes.